A Fully Automatic Procedure for Brain Tumor Segmentation from Multi-Spectral MRI Records Using Ensemble Learning and Atlas-Based Data Enhancement

Abstract

The accurate and reliable segmentation of gliomas from magnetic resonance image (MRI) data has an important role in diagnosis, intervention planning, and monitoring the tumor’s evolution during and after therapy. Segmentation has serious anatomical obstacles like the great variety of the tumor’s location, size, shape, and appearance and the modified position of normal tissues. Other phenomena like intensity inhomogeneity and the lack of standard intensity scale in MRI data represent further difficulties. This paper proposes a fully automatic brain tumor segmentation procedure that attempts to handle all the above problems. Having its foundations on the MRI data provided by the MICCAI Brain Tumor Segmentation (BraTS) Challenges, the procedure consists of three main phases. The first pre-processing phase prepares the MRI data to be suitable for supervised classification, by attempting to fix missing data, suppressing the intensity inhomogeneity, normalizing the histogram of observed data channels, generating additional morphological, gradient-based, and Gabor-wavelet features, and optionally applying atlas-based data enhancement. The second phase accomplishes the main classification process using ensembles of binary decision trees and provides an initial, intermediary labeling for each pixel of test records. The last phase reevaluates these intermediary labels using a random forest classifier, then deploys a spatial region growing-based structural validation of suspected tumors, thus achieving a high-quality final segmentation result. The accuracy of the procedure is evaluated using the multi-spectral MRI records of the BraTS 2015 and BraTS 2019 training data sets. The procedure achieves high-quality segmentation results, characterized by average Dice similarity scores of up to 86%.

Negligible effect of brain MRI data preprocessing for tumor segmentation

“…Interestingly, while the general scope of MRI preprocessing methods is set, there is no consensus regarding which of them should be used, in what configuration, and in which order. For instance, all observed papers analyzing multisequence MRI use inter-modality registration [9,20,18,14,8,7]; most of the papers use some kind of voxel resampling, either to an isotropic voxel (e.g. 1×1×1mm 3 ), or to the same image resolution (in voxels), or both, by means of non-rigid atlas registration [9,20,18,14,8,7,21,2].…”

“…For instance, all observed papers analyzing multisequence MRI use inter-modality registration [9,20,18,14,8,7]; most of the papers use some kind of voxel resampling, either to an isotropic voxel (e.g. 1×1×1mm 3 ), or to the same image resolution (in voxels), or both, by means of non-rigid atlas registration [9,20,18,14,8,7,21,2]. Two exceptions are [12], where authors train on retrospective data (with inclusion/exclusion criteria) and test on prospective data, collected with unified scanning protocol; and [24], who again, used data acquired with the same scanning protocol 6 , thus, almost no domain variability.…”

“…Two exceptions are [12], where authors train on retrospective data (with inclusion/exclusion criteria) and test on prospective data, collected with unified scanning protocol; and [24], who again, used data acquired with the same scanning protocol 6 , thus, almost no domain variability. In contrast, multi-institutional studies typically use some intensity or noise correction approaches, with most popular being: histogram matching [9,18,21], biasfield correction [9,18,7,21], denoising [18,21] and z-scoring [20,18,14,8,21,24,12]. Finally, all but one observed paper [24] use skull stripping, arguing that nonbrain tissue is a significant source of error for downstream tumor segmentation [5].…”

Comparison of Heat Demand Prediction Using Wavelet Analysis and Neural Network for a District Heating Network

“…Thanks to the work of Meyer and Mallatt [12,13], wavelets have become widely known. Wavelet transform, thanks to its properties, is usable in many fields-mainly picture and video processing [14][15][16][17], fault detection [18][19][20], diagnostics and research in medicine [21,22], but also in many other fields [23][24][25].…”

Automated Detection of Brain Tumor through Magnetic Resonance Images Using Convolutional Neural Network

“…In radiology, these methods provide a great deal of awareness of diagnosis, treatment, and perception [ 32 ]. This study is aimed at providing the solution to existing problems of segmentation and generating a high-quality outcome with less computation time and error rate, using transfer learning-based classification without the use of specialized hardware, which is not accessible in underdeveloped countries with multiple image processing tasks equipped for MRIs with focal disabilities [ 33 ]. Therefore, the developed method is efficient and reliable for the automated detection of brain tumor.…”